Date: Tuesday, May 17 (Main Conference Day 1)
Start Time: 4:50 pm
End Time: 5:20 pm
For most vision applications, the best solution combines classical vision algorithms with AI techniques. In practice, developing hybrid solutions is challenging. For example, different tools are typically used for different types of algorithms. And creating efficient implementations for embedded processors poses additional challenges, because what works in the cloud or on a PC often is not suited for an embedded target. Optimized implementations of individual algorithms are often available for certain processors, but these are not portable to other processors, and are not designed to interoperate with each other. We will introduce a flexible software ecosystem that works across various hardware target systems, providing vision system developers with a “no-code” UI and tools to streamline the implementation of image acquisition, classical vision algorithms and AI inference. This enables software providers to easily adapt their software to various deployment targets, and empowers processor manufacturers to enable their customers to develop computer vision solutions quickly.